1 Get data

1.1 Get utilities in S&P data

Original attempt was to create custom market factors for within the S&P universe. After extensive searching it was found impossible to collect the data required without paying for a service. Therefore an alternative method of evaluating the factor model was required.

Exploring the difference between Fama French factors vs Asness and Frazzini factors as explained in Asness, C.S, and Frazzini, A. (2013), The Devil in HML’s Details.

Information on paper can be found: https://www.aqr.com/Insights/Research/Journal-Article/The-Devil-in-HMLs-Details

Paper can be found: https://www.iijournalseprint.com/JPM/AQR/Sum13DevilinHMLsDetails05t/index.html

Data can be found: https://www.aqr.com/Insights/Datasets/The-Devil-in-HMLs-Details-Factors-Monthly

The Journal of Portfolio Management, Vol. 39, No. 4, pp. 49-68

1.2 Get factor data

2 Allocation Methods

2.1 Mean-Variance

2.2 Method 1: Factor Model

2.3 Method 2: James-Stein with Ledoit Wolf

2.3 Benchmark: 1/n

3 Rolling Window

4 Simulation

Things to Look into:

  1. Devil vs FF performance in 3 factor
  2. Devil vs FF performance in 5 factor (use FF data for all 4 factors and just swap in devil vs not devil)
  3. Devil vs FF performance in 6 factor (5 factor with momentum factor)
  4. What is a good rolling window to use for calculation? 5 years back? 2 years? 1 year? how does our window choice affect returns? What is the optimal Window?
    • What does this say on how long the market stays constant for?

Needed from code perspective: